Every paper, one deep dive at a time.

Biased or Personalized? The Impact of Personal Information on AI-driven Development

Billions of Sketches Reveal Hidden Cultural Variation in Human Concepts

Unicode TAG-Block Concealment of Tool-Metadata Payloads in the Model Context Protocol: An Approval-View Fidelity Gap Across Three Independent Server Implementations

More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges

Multiplayer Interactive World Models with Representation Autoencoders

Rating the Pitch, Not the Product: User Evaluations of LLMs Reflect Expectations More Than Performance

How Much is Left? LLMs Linearly Encode Their Remaining Output Length

Verbalizable Representations Form a Global Workspace in Language Models

Safety Testing LLM Agents at Scale: From Risk Discovery to Evidence-Grounded Verification

Phantom References: Hallucinated Citations That Survive Peer Review at Top-Tier Conferences

EVOCHAMBER: Test-Time Co-evolution of Multi-Agent System at Individual, Team, and Population Scales

Mechanistically Eliciting Latent Behaviors in Language Models

Can Language Models Actually Retrieve In-Context? Drowning in Documents at Million Token Scale

IsoSci: A Benchmark of Isomorphic Cross-Domain Science Problems for Evaluating Reasoning versus Knowledge Retrieval in LLMs

The Agentic Garden of Forking Paths

Coding Agents Are Guessing: Measuring Action-Boundary Violations in Underspecified DevOps Instructions

ASPIRE: Agentic /Skills Discovery for Robotics

Is One Layer Enough? Training A Single Transformer Layer Can Match Full-Parameter RL Training

AutoMem: Automated Learning of Memory as a Cognitive Skill

Modality-Driven Search with Holistic Trace Judging for ARC-AGI-2

ClawArena-Team: Benchmarking Subagent Orchestration and Dynamic Workflows in Language-Model Agents

Xiaomi-GUI-0 Technical Report

Beyond the Library: An Agentic Framework for Autoformalizing Research Mathematics

An AI agent for treatment reasoning over a biomedical tool universe

Hierarchical Experimentalist Agents

It Lied to a Doctor to Buy Poison Ingredients: Quantifying Real-World Misuse of Phone-use Agents

Tool Use Enables Undetectable Steganography in Multi-Agent LLM Systems

Internalizing the Future: A Unified Agentic Training Paradigm for World Model Planning

Grounded Iterative Language Planning: How Parameterized World Models Reduce Hallucination Propagation in LLM Agents

GBC: Gradient-Based Connections for Optimizing Multi-Agent Systems

Supersede: Diagnosing and Training the Memory-Update Gap in LLM Agents

DSpark: Confidence-Scheduled Speculative Decoding with

The Red Queen Gödel Machine: Co-Evolving Agents and Their Evaluators

Optimizing CUDA like a Human: Micro-Profiling Tools as Expert Surrogates for LLM-Based GPU Kernel Optimization

Closing the Loop to Discover Psychological Theories with an Automated Cognitive Scientist

Localizing RL-Induced Tool Use to a Single Crosscoder Feature

Model Forensics: Investigating Whether Concerning Behavior Reflects Misalignment

Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents

Cliff Tokens: Identifying Single-Token Failure Triggers in LLM Mathematical Reasoning

Do Thinking Tokens Help with Safety?

Bayesian control for coding agents

SHERLOC: Structured Diagnostic Localization for Code Repair Agents

Metis: Bridging Text and Code Memory for Self-Evolving Agents

Qwen-AgentWorld: Language World Models for General Agents

Sakana Fugu Technical Report

Group-Graph Policy Optimization for Long-Horizon Agentic Reinforcement Learning

Governance Decay: How Context Compaction Silently Erases Safety Constraints in Long-Horizon LLM Agents

Provable Benefits of RLVR over SFT for Reasoning Models: Learning to Backtrack Efficiently

Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning

Playful Agentic Robot Learning

Connect the Dots: Training LLMs for Long-Lifecycle Agents with Cross-Domain Generalization Via Reinforcement Learning

ENPIRE: Agentic Robot Policy Self-Improvement in the Real World

FloatDoor: Platform-Triggered Backdoors in LLMs

Beyond the GUI Paradigm: Do Mobile Agents Need the Phone Screen?

ProCUA-SFT Technical Report

Skill-Guided Continuation Distillation for GUI Agents

Native Active Perception as Reasoning for Omni-Modal Understanding

Rift: A Conflict Signature for Deception in Language Models

Self-CTRL: Self-Consistency Training with Reinforcement Learning

Not All Skills Help: Measuring and Repairing Agent Knowledge

CoAgent: Concurrency Control for Multi-Agent Systems

Is Your Agent Playing Dead? Deployed LLM Agents Exhibit Constraint-Evasive Fabrication and Thanatosis

Greed Is Learned: Visible Incentives as Reward-Hacking Triggers

HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry

From Shield to Target: Denial-of-Service Attacks on LLM-Based Agent Guardrails

Natively Unlearnable Large Language Models

When the Tool Decides: LLM Agents Defer Blindly to Graph Neural Network Tools, and Stronger Backbones Defer More

Prefill Awareness in Large Language Models

HarnessBridge: Learnable Bidirectional Controller for LLM Agent Harness

Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning

Beyond the Commitment Boundary: Probing Epiphenomenal Chain-of-Thought in Large Reasoning Models

Arbor: Tree Search as a Cognition Layer for Autonomous Agents

MaxProof: Scaling Mathematical Proof with Generative-Verifier RL and Population-Level Test-Time Scaling

SkillAxe: Sharpening LLM-Authored Agent Skills Through Evaluation-Guided Self-Refinement

Toward Generalist Autonomous Research via Hypothesis-Tree Refinement

Decentralized Multi-Agent Systems with Shared Context

Harnessing the Collective Intelligence of AI Agents in the Wild for New Discoveries

Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalization

TextLDM: Language Modeling with Continuous Latent Diffusion

Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills

SWE-Marathon: Can Agents Autonomously Complete Ultra-Long-Horizon Software Work?

Hardening Agent Benchmarks with Adversarial Hacker-Fixer Loops

Emergence World: A Platform for Evaluating Long-Horizon Multi-Agent Autonomy

Lean4Agent: Formal Modeling and Verification for Agent Workflow and Trajectory

From Failed Trajectories to Reliable LLM Agents: Diagnosing and Repairing Harness Flaws

Retrospective Harness Optimization: Improving LLM Agents via Self-Preference over Trajectory Rollouts

Agentic Monte Carlo: Simulating Reinforcement Learning for Black-Box Agents

Safety Paradox: How Enhanced Safety Awareness Leaves LLMs Vulnerable to Posterior Attack

Goedel-Architect: Streamlining Formal Theorem Proving with Blueprint Generation and Refinement

Streaming Communication in Multi-Agent Reasoning

MIRAGE: Mobile Agents with Implicit Reasoning and Generative World Models

Scaling Self-Evolving Agents via Parametric Memory

What If Prompt Injection Never Left? Exploring Cross-Session Stored Prompt Injection in Agentic Systems

The Meta-Agent Challenge: Are Current Agents Capable of Autonomous Agent Development?

OpenWebRL: Demystifying Online Multi-turn Reinforcement Learning for Visual Web Agents

Inducing Reasoning Primitives from Agent Traces

EvoTrainer: Co-Evolving LLM Policies and Training Harnesses for Autonomous Agentic Reinforcement Learning

The Deterministic Horizon: When Extended Reasoning Fails and Tool Delegation Becomes Necessary

Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactions

ExpGraph: Model-Agnostic Experience Learning with Graph-Structured Memory for LLM Agents

From Prompt Injection to Persistent Control: Defending Agentic Harness Against Trojan Backdoors

MosaicLeaks:Privacy Risks in Querying-in-the-Open for Deep Research Agents

Emergent Languages in Populations of Language Model Agents: From Token Efficiency to Oversight Evasion

Stateful Online Monitoring Catches Distributed Agent Attacks

Formalizing Mathematics at Scale

PokerSkill: LLMs Can Play Expert-Level Poker without Training or Solvers

Self-Trained Verification for Training- and Test-Time Self-Improvement

Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet

Scaling Laws for Agent Harnesses via Effective Feedback Compute

Shepherd: A Runtime Substrate Empowering Meta-Agents with a Formalized Execution Trace

AutoScientists: Self-Organizing Agent Teams for Long-Running Scientific Experimentation

The Fragility of Chain-of-Thought Monitoring Across Typologically Diverse Languages

Calibrating Conservatism for Scalable Oversight

LiveBrowseComp: Are Search Agents Searching, or Just Verifying What They Already Know?

Why LLMs Fail at Causal Discovery and How Interventional Agents Escape

The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence

ScientistOne: Towards Human-Level Autonomous Research via Chain-of-Evidence

SIA: Self Improving AI with Harness & Weight Updates

A Universal Cliff and a Design Fingerprint: Cross-Section Defect Detection Under LLM Orchestration

Your Agents Are Aging Too: Agent Lifespan Engineering for Deployed Systems

Language Models Need Sleep

ECHO: Terminal Agents Learn World Models for Free

AgentFugue: Agent Scaling for Long-Horizon Tasks through Collective Reasoning

QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks

Understanding and Mitigating Premature Confidence for Better LLM Reasoning

CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents

Metacognition as Reward: Reinforcing LLM Reasoning via Knowledge and Regulation Signals

SkillOpt: Executive Strategy for Self-Evolving Agent Skills

When Do LLMs Reason? A Dynamical Systems View via Entropy Phase Transitions

RMA: an Agentic System for Research-Level Mathematical Problems

Inductive Deductive Synthesis: Enabling AI to Generate Formally Verified Systems

HRM-Text: Efficient Pretraining Beyond Scaling

Multi-LLM Systems Exhibit Robust Semantic Collapse

Qumus: Realization of An Embodied AI Quantum Material Experimentalist

Adapting the Interface, Not the Model: Runtime Harness Adaptation for Deterministic LLM Agents

Hallucination as Commitment Failure: Larger LLMs Misfire Despite Knowing the Answer

Is Capability a Liability? More Capable Language Models Make Worse Forecasts When It Matters Most

DeltaBox: Scaling Stateful AI Agents with Millisecond-Level Sandbox Checkpoint/Rollback

Advancing Mathematics Research with AI-Driven Formal Proof Search

ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents

optimize_anything: A Universal API for Optimizing any Text Parameter

Auto-Dreamer: Learning Offline Memory Consolidation for Language Agents

Agent JIT Compilation for Latency-Optimizing Web Agent Planning and Scheduling

Hallucination as Exploit: Evidence-Carrying Multimodal Agents

Agent Meltdowns: The Road to Hell Is Paved with Helpful Agents

NeuroMAS: Multi-Agent Systems as Neural Networks with Joint Reinforcement Learning

Firefly: Illuminating Large-Scale Verified Tool-Call Data Generation from Real APIs

The Capability Paradox: How Smarter Auditors Make Multi-Agent Systems Less Secure

ADR: An Agentic Detection System for Enterprise Agentic AI Security

Judge Circuits

Training on Documents About Monitoring Leads to CoT Obfuscation

Agentic Discovery of Neural Architectures: AIRA-Compose and AIRA-Design

Look Before You Leap: Autonomous Exploration for LLM Agents

Argus: Evidence Assembly for Scalable Deep Research Agents

Ambient Persuasion in a Deployed AI Agent: Unauthorized Escalation Following Routine Non-Adversarial Content Exposure

Achieving Gold-Medal-Level Olympiad Reasoning via Simple and Unified Scaling

Orchard: An Open-Source Agentic Modeling Framework

Harnessing Agentic Evolution

LLM-Based Persuasion Enables Guardrail Override in Frontier LLMs

History Anchors: How Prior Behavior Steers LLM Decisions Toward Unsafe Actions

Negation Neglect: When models fail to learn negations in training

GRAFT-ATHENA: Self-Improving Agentic Teams for Autonomous Discovery and Evolutionary Numerical Algorithms

Solve the Loop: Attractor Models for Language and Reasoning

The Bicameral Model: Bidirectional Hidden-State Coupling Between Parallel Language Models

Oracle Poisoning: Corrupting Knowledge Graphs to Weaponise AI Agent Reasoning

How LLMs Are Persuaded: A Few Attention Heads, Rerouted

The Geometry of Forgetting: Temporal Knowledge Drift as an Independent Axis in LLM Representations

Sparse Attention as a Range Searching Problem: Towards an Inference-Efficient Index for KV Cache

Ask Early, Ask Late, Ask Right: When Does Clarification Timing Matter for Long-Horizon Agents?

TraceFix: Repairing Agent Coordination Protocols with TLA+ Counterexamples

Echo: KV-Cache-Free Associative Recall with Spectral Koopman Operators

State Stream Transformer (SST) V2: Parallel Training of Nonlinear Recurrence for Latent Space Reasoning

STALE: Can LLM Agents Know When Their Memories Are No Longer Valid?

LoopTrap: Termination Poisoning Attacks on LLM Agents

AI Co-Mathematician: Accelerating Mathematicians with Agentic AI

Recursive Agent Optimization

VibeServe: Can AI Agents Build Bespoke LLM Serving Systems?

Rethinking RL for LLM Reasoning: It's Sparse Policy Selection, Not Capability Learning

Explaining and Preventing Alignment Collapse in Iterative RLHF

Agentic Vulnerability Reasoning on Windows COM Binaries

What Happens Inside Agent Memory? Circuit Analysis from Emergence to Diagnosis

Model Spec Midtraining: Improving How Alignment Training Generalizes

OpenSeeker-v2: Pushing the Limits of Search Agents with Informative and High-Difficulty Trajectories

The Compliance Gap: Why AI Systems Promise to Follow Process Instructions but Don't

EvoLM: Self-Evolving Language Models through Co-Evolved Discriminative Rubrics

What Suppresses Nash Equilibrium Play in Large Language Models? Mechanistic Evidence and Causal Control

Gym-Anything: Turn any Software into an Agent Environment

MARS: Efficient, Adaptive Co-Scheduling for Heterogeneous Agentic Systems

Political Bias Audits of LLMs Capture Sycophancy to the Inferred Auditor

Guiding Symbolic Execution with Static Analysis and LLMs for Vulnerability Discovery

Why Search When You Can Transfer? Amortized Agentic Workflow Design from Structural Priors

Dynamic analysis enhances issue resolution

Does RL Expand the Capability Boundary of LLM Agents? A PASS@(k,T) Analysis

RAGEN-2: Reasoning Collapse in Agentic RL

SFT-then-RL Outperforms Mixed-Policy Methods for LLM Reasoning

A Subgoal-driven Framework for Improving Long-Horizon LLM Agents

Exploration Hacking: Can LLMs Learn to Resist RL Training?

Emotion Concepts and their Function in a Large Language Model

Empowering Autonomous Debugging Agents with Efficient Dynamic Analysis

LLMs Know They're Wrong and Agree Anyway: The Shared Sycophancy-Lying Circuit

Scaling Test-Time Compute for Agentic Coding

End-to-end autonomous scientific discovery on a real optical platform

Peer-Preservation in Frontier Models